Speech Accent Archive website and listened carefully to their recordings of six Albanians. The accent was more subtle than I had expected but still intriguing. Generalizations were offered:I auditioned for a TV pilot the other day (think The Sopranos with Albanians in cottage country—comedy), and had to quickly pick up an accent. I visited the
- ‘r’ to trill (e.g., flipped or slightly trilled ‘r’),
- ‘w’ to labial fricative (“Wednesday” becomes “Vednesday”),
- non-aspiration (“please” has no breathy noise on the ‘p’),
- final obstruent devoicing (“Bob” becomes “Bop”),
- interdental fricative to stop (“the” becomes “de”),
- vowel shortening, raising and retracting.
The people at my audition said my accent was excellent . . . but they hinted that I didn’t look the part—an avuncular assassin from the Balkans. (No, I look more like a music theory professor from an under-funded university. Without tenure.)
Are accents of interest to composers and sound designers? Accents involve interpretations of consonants and vowels, i.e., how something is said by members of different groups. Saying the “ay” vowel (or dipthong) in the word “saying” will be done differently by Americans and Australians. It’s still written as “ay” but there are differences in the way the vocal tracts are made to resonate by the person speaking.
Vowel raising means that vowels become slightly more closed, e.g., “sit” sounds like “seat,” by displacing the tongue nearer to the roof of the mouth, while vowel retracting means to displace the tongue backwards and thereby narrow the pharynx, e.g., “slab” leans a little toward the sound of “slob.” You can visualize these shifts on a vowel diagram.
Acoustically this is achieved in a male voice by changing the first formant, or region of resonance, from 400 Hz down to 275 Hz and the second formant slightly upward from 2000 Hz to 2300 Hz. You could apply these resonance peaks to white noise and it would sound like two similar yet distinct vowels: resonant centre frequencies of 400 & 2000 Hz (“ih”) versus 275 & 2300 Hz (“ee”).
Sometimes a chord will take on the character of a vowel, e.g., in Bach’s Brandenburg Concerto No. 5 in G, second movement (Allegro), measure 11. No matter if it’s played by a string orchestra or on a Moog synthesizer I always hear a choir singing, “La la la” on the A dominant 7th, 3rd inversion. The illusion does not work in the penultimate bar when the theme has been recapitulated in the home key. The corresponding measure is transposed a fifth lower but does not sound like “la” because it no longer shares resonance characteristics with a male singing the vowel “ah.”
Interested in vocal synthesis, I visited Paris’s IRCAM on June 19 during a vacation in France. Xavier Rodet gave a public demonstration of vocal synthesis and transformation software they are developing which is of interest to the film industry: SuperVP-TRaX, and SuperVP-VoiceForger. Not only can they change the perceived age of a speaker by changing the frequency, resonances and timbre, e.g., make a boy sound like an infirm, old woman, but they can now impose a cultural accent on the speaker.
He played a short excerpt of an actor speaking with his Parisian accent, then Rodet turned some dials, and the actor now sounded like he had a Québecois accent. (The audience chuckled. Are they tickled by a Québecois accent like we’re tickled by a Scottish accent?)
He played another short example: from the film Vatel Gérard Depardieu said something in English with his normal French accent. Suppose, however, the producer finds the actor’s delivery too flat for the target audience, i.e., not French-sounding enough to Americans’ ears. Instead of hiring the expensive and busy actor to re-record his lines in a studio, then synchronize the new lines to the picture, they can now turn up the ‘French’ dial and simply exaggerate his French accent.
There are courses available to people for changing their accent, usually to reduce their own accent and learn the local one, especially useful to actors and immigrants. I would think this kind of software could be used to help with the training by listening, analyzing and indicating weak areas in the speaker’s development.
I think accent-changing software could also be used by telemarketers to match their accent with that of their target market. No longer would a telemarketer from India be limited to weather reports and local news from Arizona in an effort to appear less foreign, they could now sound like they were Arizonians themselves, at the flick of a switch, without having to take any elocution training. Or the translation software could identify each particular call recipient’s accent by analysing his or her voice, and subtly changing the telemarketer’s accent to match the listener’s, regardless of what city the call recipient lives in.
Furthermore, I imagine the accent and voice transformation software could be used deceptively to make someone’s entire voice sound like that of some other particular person, e.g., to make my voice sound like . . . the Prime Minister’s! I see doors opening for pranksters and spies alike. When can I download such an app to my smartphone?